This project investigates a novel treatment for word retrieval problems in aphasia, a debilitating language disorder typically resulting from a stroke on the left side of the brain. In the typical adult system, the ability to think of words is suppored by a complex set of processes that operate automatically and outside of conscious control (i.e., implicitly). Prior research evidence suggests that these automatic processes are impaired in aphasia, operating on a different time-course than in the typical system. The single-subject, multiple baseline design study proposed in this application will assess a treatment designed to directly address these automatic processes in order to restore rapid, accurate language processing. This treatment has already been tested in a Phase 1 study with two individuals with aphasia. Results demonstrated that naming became more accurate for treated items but substantially less so for untrained items within the same categories. In addition, some cross- category generalization was evident. The early Phase 2 study proposed here is the next critical step in developing this novel treatment methodology, assessing its effects in 10 individuals with aphasia. The aims of this study are to assess treatment, generalization, and maintenance effects of this treatment approach. The treatment being studied in this project involves presenting picture names subliminally (masked priming), to encourage automatic processing without engaging explicit, conscious processes that could interfere with the automatic response. The masked primes will be repeatedly paired with their respective pictures, in order to rebuild the connections between words and their referents within the language processing network. After multiple pairings of the masked primes and pictures, the participants will be asked to name the picture, encouraging access to the portions of the language network pre-stimulated by the prime-picture pairs. This procedure will be repeated 4 times per session for each item being trained. Two semantic categories will be trained for 16 sessions each. Probe tests of naming accuracy and response time will be conducted multiple times before, during, and after treatment to assess cumulative changes in function as a result of the treatment. In addition, within-category and across-category generalization will be measured by repeated naming probes of untrained items. Findings from this study will be used to pursue larger-scale development of treatment methods for anomia that directly target restoration of automatic language processing networks.